Glossary
Retrieval-augmented generation
Answering by first retrieving passages from a known collection, then writing from those passages.
A model that answers only from its training will not know your notes. Retrieval-augmented generation, often called RAG, searches a collection you supply and conditions the answer on what came back. The retrieval step is what makes the answer about your files instead of about the public web.
Cairn indexes your files and connected tools, then answers with a citation on every sentence. That is the shape of this pattern as the product describes it. We do not name a model vendor or a vector database on this site. The behavior to test is the citation, which is citation grounding.